Graph-augmented LLMs for social media-based clinical trial recruitment

Clinical trials (CTs) are essential for advancing disease diagnosis and treatment; however, recruiting eligible participants remains a major bottleneck. Existing recruitment strategies, such as advertisements and electronic health record-based screening, are often time-consuming and inefficient. Social networks provide a promising alternative source for identifying potential participants, but this direction remains underexplored because user information is sparse and current methods make limited use of interaction structure.

To address these challenges, we propose a framework that predicts whether a social media user is a potential candidate for CTs. The framework uses summarization to organize fragmented posts and a two-stage graph-based procedure that treats neighboring users' initial labels as contextual cues rather than deterministic labels.

Experiments conducted on two Reddit clinical trial datasets show that GraphSum improves performance in most evaluated settings. Ablation results indicate that the effects of summarization, explanation of medical terms, and graph refinement vary across models and datasets.

Integrating summarization with two-stage graph-based modeling can improve the identification of potential CT participants from social media, although the benefit depends on the model, dataset, and eligibility criterion.